chore: launch-hygiene pass — advisory IDs, README register, advisory registry, language bar#3
Merged
Merged
Conversation
Self-found project issues were labeled CVE-2026-001/002/003. CVEs are issued by CNAs, not projects; the labels read as inflation to a security-literate audience. Rename to internal advisory IDs (JG-ADV-2026-001/002/003, numeric mapping preserved) across all docs, and add a one-line disclaimer at first use in README.md and THREAT_MODEL.md. Docs only; no code/logic touched. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Title said 'Enterprise Semantic Firewall' while the validation-status section honestly states 'not independently audited or enterprise-GA / research prototype.' That register collision undercuts the project's core intellectual-honesty asset. Retitle to match what the validation section already claims. The Fleet & Enterprise feature-tier section (feature-gated, off by default) is legitimate and left as-is. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Linguist reports ~83% C because of the bpf/*.bpf.c / bpf/lsm/*.c objects, but the project is pitched and built as Rust. Vendor the whole bpf/ tree for language-stat purposes (source stays visible; only Linguist stats change). GitHub re-indexes the language bar on the next push. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Records Tasks A-D (done) and the remaining MANUAL items: re-record demo without recorder overlay, post as native LinkedIn video with repo URL in first comment, warm DM for AFWERX/DIU-adjacent repost. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…llision) JG-ADV-2026-003 ambiguously referred to two findings: 'Agent Impersonation via UID Spoofing' (red-team-report.md, disclosed 2026-06-08) and 'socket-LSM fail-open' (THREAT_MODEL/BENCHMARKS-04, surfaced 2026-06-14, Run 04). The newer finding takes the new number: the socket fail-open becomes JG-ADV-2026-004; UID-spoofing keeps JG-ADV-2026-003. CHANGELOG.md has no 003 reference; red-team-report.md already holds 003 for UID-spoofing — both unchanged. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Single source of truth for JG-ADV-* IDs: ID | Title | Component | Disclosed | Status | Fix commit, one row per advisory (001-004), with per-advisory notes. README Known Limitations now points to it. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…ings Comment/echo/printf strings only — no logic changed. Maps to the renumbered scheme: interpreter -> JG-ADV-2026-001, path-resolution -> JG-ADV-2026-002, socket fail-open -> JG-ADV-2026-004. Verified no test or script asserts on these literals, so no output contracts break. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Replace 'guarantee absolute zero-trust process isolation and immutable anti-replay protection across the entire host subsystem' with 'enforce zero-trust process isolation and anti-replay protection for governed cgroups' — aligns with the cgroup-scoped enforcement described in the validation-status and Known Limitations sections. No new claims. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Show the CVE -> JG-ADV mappings as a rename record, including the 003 collision split (socket fail-open renumbered to JG-ADV-2026-004), and point to SECURITY/ADVISORIES.md as the canonical registry. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- README: drop inaccurate 'on AlphaOS' (validated on Debian/Ubuntu/AlmaLinux) - THREAT_MODEL CVE log: reconcile JG-ADV-2026-004 status to 'Fixed (re-validated on AlmaLinux 9 / 5.14, Run 04)' per BENCHMARKS-04 §2 - THREAT_MODEL CVE log: add the missing JG-ADV-2026-003 (UID spoofing) row for parity with SECURITY/ADVISORIES.md - re-align the ASCII box border in validate_m2_path_resolution.sh after the longer JG-ADV identifier (cosmetic) Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Full automated suite + userspace benchmarks re-run on the branch after the docs/comment-only hygiene pass, on the same CPU/distro/kernel as Run 01 (Ryzen 5 7520U / Debian 13 / 6.12). 116/116 green, 0 fail-open in the adversarial suite, P50 259us (vs Run 01's 257us), peak ~6,208 RPS, 0 errors. Kernel Tier 4 not run on this unprivileged sandbox (see Runs 01-04). Confirms the hygiene changes altered nothing operational. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…C/Rust language-bar note; align qualifier to validation section Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Launch-hygiene pass (docs + metadata only — no logic)
Bounded hygiene pass ahead of the first distribution push to a security-literate
audience. Closes two credibility landmines, adds an advisory registry, and
re-validates the suite. No Rust/C logic, Cargo, policy, benchmark numbers, or
validation-section substance changed.
What changed
CVE-2026-00x→JG-ADV-2026-00xacross all docs and code comments / script outputstrings (CVEs are issued by CNAs, not projects). One-line disclaimer added at
first use in
README.mdandTHREAT_MODEL.md.JG-ADV-2026-003had referred to two findings.Disclosure dates settle it: UID-spoofing (2026-06-08) keeps
-003; the newersocket-LSM fail-open (2026-06-14, Run 04) becomes
JG-ADV-2026-004.SECURITY/ADVISORIES.md(ID · Title · Component ·Disclosed · Status · Fix commit) — single source of truth, linked from README.
enforcement firewall for autonomous AI agents (research prototype)" to match
the validation-status section; softened the line-7 overclaim to defensible,
cgroup-scoped wording. (The legitimate feature-gated Fleet & Enterprise tier
section is untouched.)
bpf/** linguist-vendoredso GitHub reads the repo asRust-primary (it re-indexes on push).
LAUNCH_CHECKLIST.md(with the explicitrename record and remaining MANUAL items); added the missing
-003row to theTHREAT_MODEL CVE log; reconciled the
-004status to "Fixed (re-validated,Run 04)".
Re-validation —
BENCHMARKS-05.mdFull userspace suite re-run on the same CPU/distro/kernel as Run 01
(Ryzen 5 7520U / Debian 13 / 6.12):
Confirms the docs/comment-only pass changed nothing operational.
Follow-up (not in this PR)
ghwas unavailable in the working environment —run the
gh repo edit … --description … --add-topic …command fromLAUNCH_CHECKLIST.mdnotes to set About/topics.LAUNCH_CHECKLIST.md: re-record demo withoutrecorder overlay; post as native LinkedIn video; warm DM for repost.
🤖 Generated with Claude Code